Mistral Vibe: Model Reasoning Shown by Default as a Collapsible Thought Row
Mistral VibeMistral Vibe 3.0.2 now shows the model's reasoning by default as a collapsible "Thought" row in the conversation. Users no longer have to switch anything on to see how the model approached a step, and they can collapse the row to keep the transcript tidy.
Key Takeaways
- Reasoning is now on by default, so users upgrading to 3.0.2 will see model thinking without changing any setting.
- It appears as a collapsible "Thought" row, which keeps long sessions readable.
- Visible reasoning helps users catch a misread task early, before edits or commands run.
- The change affects everyone who upgrades, because it alters a default rather than adding an opt-in toggle.
- Collapsing the row hides the detail without losing it, so it stays available when needed.
- Together with the stricter subagent rules in the same release, 3.0.2 leans toward transparency and control over the agent's behavior.
Reasoning is visible out of the box
Mistral Vibe 3.0.2 changes a default. The model's reasoning now appears in the session by default, as a collapsible row labeled "Thought". Previously this view was not shown unless enabled, so many users never saw it.
What users will notice
When the model reasons before answering or calling a tool, that reasoning appears in the transcript as its own row. The row can be collapsed, so users who prefer a compact view can fold it away. Users who want to follow or audit the agent's thinking can leave it open.
Why it matters
Seeing the reasoning makes it easier to spot when an agent misread a task before it edits files or runs commands. It also helps when explaining or reviewing what a session did. Because the row is collapsible, the cost for people who do not want it is one click.
This is a change to existing behavior rather than a new opt-in feature. Anyone upgrading to 3.0.2 will see Thought rows appear in sessions where they previously saw none.